Statistical inference for Markov chains with lumped states
Ahmed Shaikh · Biometrika · 1976
In an ( n +1)-state homogeneous Markov chain, when due to some experimental limitations n states are not observable, the maximum likelihood estimation of an unknown parameter θ is considered and the efficiency of θ relative to the most efficient estimator when all the states are observable is discussed. The study is extended to the case when there are two lumps of unobservable states in a four-state Markov chain.